BEP-1: Short term plasticity
Abstract: Implementation of short term plasticity following the Tsodyks-Markram
model.

The model: presynaptic spikes produce instantaneous changes in synaptic
variables, which also follow differential equations. The synaptic weight is
modulated by one of those variables, in a multiplicative way.

References
==========
Mongillo et al (2005). Synaptic Theory of Working Memory, Science.
dx/dt=(1-x)/taud  (depression)
du/dt=(U-u)/tauf  (facilitation)
spike: u->u+U*(1-u); x->x*(1-u)
u*x is the modulation factor (in 0..1), weights are modulated by the value of
u*x before updating.

Markram et al (1998). Differential signaling via the same axon of neocortical
pyramidal neurons, PNAS.
[same model]

Tsodyks & Markram (1997). The neural code between neocortical pyramidal neurons
depends on neurotransmitter release probability, PNAS.

Senn et al (2001). An Algorithm for Modifying Neurotransmitter Release
Probability Based on Pre- and Postsynaptic Spike Timing, Neural Comp.

Implementation
==============
A number of variables are involved in the model. Since they are only modified
by presynaptic spikes, a group of the size of the presynaptic group P should be
created, with the appropriate dynamics. This group G receives spikes from P through
a SpikeMonitor, which modify its variables in an event-driven way.
Spikes are directly forwarded to the postsynaptic group Q, with a modulation.
That connection should replace the original connection between P and Q.
This can be done simply by replacing the source variable of the Connection.

To deal with delays correctly, the delays from P->Q should be transferred to
P->G and removed from P->Q.

The value of the modulation variable is only required at presynaptic spike times.
It implies that the integration of the equations can be event-driven, which would
minimize the simulation cost.

Interface
=========
A similar interface as for STDP should be chosen, e.g.:
stp=STP(C,parameters)
where C is the Connection.

One could pass the equations as a string as for STDP, but are there any other
models of synaptic dynamics being used?
